Software‐defined network provides eminent solutions for many complex network management functionalities in a data center network (DCN). One of the major tasks in any network is the load balancing in the available links. Due to dynamic data traffic nature in network, it is necessary to perform deep learning approaches for the long‐term and the short‐term data. This paper proposes a splitting policy‐based RL network (SPRLN) approach, a reinforcement learning‐based proactive load balancing algorithm that avoids the poll to the controller after the switch encounters an abnormality. The proposed method has been tested with simulations and found successful in improving the overall network performance by taking appropriate action for reward maximization. The testbed environment is treated as a Q‐learning algorithm; here, the optimality is defined as the path having the least score so that the overloaded path in that particular time can be avoided. An artificial neural network is needed because the data are uncertain all the time. Thus, the proposed SPRLN method yields 30% increased throughput and with 80% reduced data loss, when compared to existing approaches.

中文翻译：

---

支持智能的方法可在软件定义的数据中心网络中实现负载平衡

软件定义的网络为数据中心网络（DCN）中的许多复杂的网络管理功能提供了出色的解决方案。任何网络中的主要任务之一是可用链路中的负载平衡。由于网络中动态数据流量的性质，有必要对长期和短期数据执行深度学习方法。本文提出了一种基于分裂策略的RL网络（SPRLN）方法，这是一种基于强化学习的主动负载均衡算法，可避免在交换机遇到异常后轮询控制器。所提出的方法已通过仿真进行了测试，发现通过采取适当的行动来最大化奖励，从而成功地改善了整体网络性能。测试平台环境被视为Q学习算法。这里，最优性定义为得分最低的路径，从而可以避免在该特定时间过载的路径。由于数据一直都是不确定的，因此需要一个人工神经网络。因此，与现有方法相比，提出的SPRLN方法可将吞吐量提高30％，数据丢失减少80％。